Study Design Considerations for Adaptive Platform Trials

ICPE 2023 Workshop on Pragmatic Trials

Jay Park

Core Clinical Sciences & McMaster University

August 23, 2023

Disclosure

  • No funding was received for preparation of this course

  • I am currently employed by Core Clinical Sciences that provides research consulting services in clinical trial designs and evidence synthesis

Intended Learning Objectives

  1. To establish common terminologies:

    • Adaptive trial designs

    • Master protocols

    • Platform trials

  2. To discuss key design considerations of adaptive platform trials

  3. To discuss a case study on adaptive platform trial for COVID-19 therapeutic outpatient research

    • The TOGETHER Trial
  4. To leave the audience with some recommendations

Conventional Trial Designs (Fixed Sample Size Designs)

  • When we think of clinical trials, we mostly imagine “one-shot” trials

\(~\)

  • A 2-arm clinical trial with a fixed sample size and one final analysis at the end

\(~\)

flowchart LR 
   A(Design) --> B(Conduct)
   B --> C(Analysis)

Main Challenge with the Conventional Approach

  • There are many unknowns. It is extremely difficult to guess right or at least uncomfortable making the guesses

\(~\)

  • In conventional trials, we only get one guess

    • If you can predict the future, no problem with the conventional approach

\(~\)

  • How do we plan clinical trials when we don’t know much about what we are studying? (e.g., COVID-19 at the start of 2020)

Adaptive Trial Designs

  • The term, adaptive trial designs is an umbrella term that refers to a group of clinical trial designs that offer pre-planned opportunity to modify aspects of an ongoing trial based on accumulating trial data

\(~\)

  • The unifying property of adaptive trial designs:

    • Use of accumulating interim data based on pre-specified plans that are developed and outlined a priori

\(~\)

  • The main motivation for adaptive trial designs is to learn from the data as they are collected during the trial and act accordingly

Comparison vs Conventional Trial Designs

  • We conduct one or more of the planned interim analyses according to the plan developed during the design stage

Sequential Designs

  • Refer to a type of trial designs that allow you to stop enrollment early (most common)

  • You can decide to allow for early stopping based on superiority and/or futility

  • Superiority: There is overwhelming evidence that the treatment works

  • Futility: There is underwhelming evidence for treatment

Motivation for sequential designs

  • Fail faster, succeed faster

Platform Trials

  • The term refers to clinical trials designed with flexibility of adding new interventions

\(~\)

  • Interventions can enter and leave at different times

\(~\)

  • They use a series of documents referred to as “master protocols” that outline common plans for evaluation of multiple interventions

Platform trials + adaptive trial designs = Adaptive platform trials

Platform trials + conventional (fixed) trial designs = Conventional (non-adaptive) platform trials

Illustration of Platform Trials

Airport Analogy

Platform Trial Design Considerations

Key Design Considerations for Platform Trials

  • Active interventions and control group

  • Allocation ratio

  • Interim analysis plans

  • Scientific merits for adding new interventions

  • Control of information flow

  • Timing of adding new interventions

Active Interventions and Control Group

  • While the number of active interventions will vary over time, platform trials should have a pre-determined maximum number of arms that can be active at a given time

\(~\)

  • Operational and feasibility considerations made to determine this number. Too many arms at once makes it very difficult

\(~\)

  • Pairwise comparisons against the common control may be preferable rather than testing for a global hypothesis testing

Allocation Ratio

  • Given the multi-arm aspects of platform trials, the probability of being randomized to the control arm can be reduced in platform trials

\(~\)

  • It is important to consider the allocation ratio for the control vs the experimental intervention(s) using simulations

\(~\)

  • Since we are comparing multiple treatments against the common control, we need adequate control data

Interim Analysis Plans

Plans for interim analyses should consider:

  • When will the first interim analysis occur (burn-in period)?

  • How many interim analyses will we conduct? And how frequently?

  • What adaptations will be allowed? What are the decision criteria?*

In addition to these statistical rules, we specify plans to prevent operational biases

  • Who will conduct the analyses? Who will be blinded and who will not be?

No one-size-fits-all solution except that

  • We should use simulation-guided design to minimize anticipated regrets

Benefits of Clinical Trial Simulations

  • Useful for planning since they allow evaluation of multiple potential scenarios and candidate designs

    • We can use simulations to compare a fixed trial design option to different adaptive trial designs

\(~\)

  • With many unknowns and assumptions that need to be made at the trial planning stage, clinical trial simulations can help to avoid trial design decisions that trial investigators would regret later after the trial shows negative findings (areas of anticipated regret)

Scientific Merits for Adding New Interventions

  • In addition to statistical considerations, scientific considerations that need to be made for adding new arms

\(~\)

  • Better to pre-specify the scientific criteria that will be used to what interventions will be added before an industry partner shows interest in participating

\(~\)

  • These scientific merits can be reviewed by an independent scientific working group that includes patient advocacy groups and other scientific stakeholders not directly involved in the trial

Scientific Merits for Adding New Interventions

Criteria used in STAMPEDE

  • Biological plausibility

  • Evidence of potential efficacy

  • Industry partnership

  • Adding a new arm would not jeopardize recruitment to the ongoing research arms

Control of Information Flow

  • Operational biases can occur if information from ongoing trial (information leakage) causes changes to participant pool, investigator behavior, and other aspects of the trial

Not new to platform trials

  • But establishing measures with strict control of communication and information flow should be implemented to prevent operational biases

  • Given perpetual nature of platform trials, information leakage may be unavoidable. Conference presentations and publications will likely become available while other interventions are being evaluated.

  • Rules to the trial are pre-specified. If rules need to be changed, the decision to change the rules should be driven by the science

Timing of Adding New Interventions

  • Concurrent control data: Data from those randomized to control during the same time as the intervention

  • Past control data (non-concurrent control): Data from those enrolled before the given intervention became active

Question:

  • Should the non-concurrent control be included into the comparison of Arm 3?

  • The correct answer is it depends

Past Control Data vs Other Historical Data

Important to note how non-concurrent data is different historical data

  • Historical RWD are prone to systematic biases since randomization is not used in real world

  • Historical randomized trial data less prone to these biases, but still study-to-study and temporal variabilities to concurrent trial

  • In platform trial, temporal variabilities still exist in past control data but maybe no (or less) study-to-study variabilities since we are using the same master protocol

Case Study: The TOGETHER Trial

The TOGETHER Trial

  • A Bayesian adaptive, placebo-controlled, platform trial that evaluated therapies for COVID-19 therapies in an outpatient setting for those at high risk of disease progression

  • Started in June 2020 with hydroxychloroquine (HCQ) and lopinavir/ritonavir (LPV/r) vs placebo

  • Has enrolled over 8,000 patients and evaluated 10+ therapies

  • Each intervention was compared against the concurrent common control

The TOGETHER Trial - Continued.

Statistical Design

  • It started with composite of hospitalization + mortality status at day 28

  • Primary endpoint did changed over time (emergency room use added after) as the disease epidemiology changed over time

  • Before new interventions were added into the platform, simulations were performed to calibrate decision rules and characterize the operating characteristics

  • Simulations (n=200,000) used to come up with stopping rules that could achieve 0.025 type I error rate (one-sided)

  • We used Bayesian beta-binomial model with non-informative priors

  • Analyses planned at 25%, 50%, 75%, and 100% of maximum sample N.

Simulation Overview

Superiority threshold:

  • >0.99 posterior probability of RR <1.00

Futility thresholds:

  • <0.20, <0.40, and <0.60

  • Gradually became more stringent over time

Simulation Overview - Continued

  • This figure is showing the posterior probability being updated at each interim analysis

  • Red shows our specified non-informative prior

  • In Arm 3, please note no Purple (3rd interim analysis) nor Orange (4th and final interim analysis), since in this virtual trial, this arm stopped for futility at Green(2nd interim analysis)

Notable Findings from the TOGETHER Trial

Fluvoxamine

Notable Findings from the TOGETHER Trial

Pegylated Interferon Lambda - A single subcutaneous injection

Conclusion and Last Remarks

The Model of Platform Trials

  • Certainly not easy, but not impossible!

Longer set-up and initial cost for platform trials

  • Trial simulation required to evaluate operating characteristics

  • Several logistical and operational considerations required

In the long-run, it’s more efficient and time/cost saving

  • Sample size savings from having a common control arm

  • Redundancies in trial set-up and close out avoided, etc

Recommendations

  • In my opinion, there is a tendency to over-complicate things

  • We don’t have to complicate things because other platform trials were complicated

Every platform trial is different

  • Multiple ways to build and maintain an “airport”

  • Let’s be honest about different trade-offs

  • The main statistical efficiency comes from multi-arm aspect of platform trial

  • We need to work in a cross-functional team, and we need to advocate for structural changes

Not easy but not impossible

  • Especially not easy the first time, it does get easier (I think?)

References

1. Park JJ, Detry MA, Murthy S, Guyatt G, Mills EJ. How to use and interpret the results of a platform trial: users’ guide to the medical literature. Jama. 2022 Jan 4;327(1):67-74.

2. Pallmann P, Bedding AW, Choodari-Oskooei B, Dimairo M, Flight L, Hampson LV, Holmes J, Mander AP, Odondi LO, Sydes MR, Villar SS. Adaptive designs in clinical trials: why use them, and how to run and report them. BMC medicine. 2018 Dec;16(1):1-5

3. Thorlund K, Haggstrom J, Park JJ, Mills EJ. Key design considerations for adaptive clinical trials: a primer for clinicians. BMJ. 2018 Mar 8;360

4. A Practical Adaptive & Novel Designs and Analysis (PANDAS) toolkit

5. Park JJ, Mills EJ, Wathen JK. Introduction to Adaptive Trial Designs and Master Protocols. Cambridge University Press; 2023 Apr 6

6. ClinicalTrials.Gov for the TOGETHER Trial: NCT04727424

7. Forrest JI et al., Resilient Clinical Trial Infrastructure in Response to the COVID-19 Pandemic: Lessons Learned from the TOGETHER Randomized Platform Clinical Trial. The American Journal of Tropical Medicine and Hygiene. 2022 Feb;106(2):389.

Book